Currently I Already did my work to transfer image on GPU and CPU. Apart from that, i try to expand the work with FPGA. Unfortunately, i still searching the suitable method how communicate CPU and FPGA (Heterogeneous Computing (FPGA-CPU))
You can transfer data by implementing a serial UART communication port on the intended FPGA. This data can be stored in a memory and input for processing in the FPGA processing building blocks. You can use the Chipscope also to input test data and output responses of your design. The details can be found by searching the web.
I think, you need create a C-code for direct memory access. This access must have link to memory who work with GPU. After this, you can use memory pointer for retranslate your data from GPU to external interface. I thinkm better use fast-interface (not UART-based)